How Children Design and Reason about Trustworthy AI Chatbots

Children's understanding of trustworthiness evolves significantly with age: Younger students (ages 10-13) equated trust with whether a chatbot fulfilled its intended purpose, while older students (ages 14-18) linked trustworthiness to transparent and calibrated design—suggesting that AI literacy programs should be developmentally tailored rather than one-size-fits-all.

Younger children may build AI systems they don't fully understand as trustworthy: Some younger students deliberately configured chatbots to give wrong answers but still labeled them as trustworthy, revealing a potential gap between children's design intentions and their grasp of what makes AI systems reliable—a critical concern as children gain access to AI-building tools.

Context matters in how children design for trust: Students calibrated academic chatbots to be significantly more transparent and formal than hobby chatbots, demonstrating that children can recognize different trust requirements across contexts, which could be leveraged in educational AI literacy curricula.

Confidence levels don't always correlate with actual trustworthiness: Younger students set significantly higher confidence levels in their chatbots than older students, suggesting younger learners may not yet understand the relationship between a system's stated confidence and its actual reliability—an important consideration for how AI systems present themselves to young users.

Seven distinct design dimensions shape children's trust perceptions: The research identified specific factors children consider when building trustworthy chatbots, providing a concrete framework for developing AI literacy tools that help children think critically about what makes AI systems worthy of trust.